MLS-C01 Question 116
Select 3You are building a predictive model to forecast housing prices using a dataset that includes features such as square footage, number of bedrooms, and proximity to schools. During the exploratory data analysis (EDA) phase, you calculate the correlation matrix and notice that the 'square footage' feature has a very high positive correlation with the target variable (housing price), while the 'proximity to schools' feature has a very low positive correlation. Based on this information, which of the following actions are appropriate when selecting features for your model?
- A
Include 'square footage' in the model as it has a strong relationship with the target variable.
- B
Exclude 'square footage' from the model because high correlation can lead to overfitting.
- C
Consider excluding 'proximity to schools' from the model as it has a low correlation with the target variable.
- D
Calculate p-values for the features to better understand their statistical significance before making final decisions.
- E
Drop both 'square footage' and 'proximity to schools' as neither guarantees better predictive performance.
Show answer and explanation
Correct answers: A, C, D
Explanation
Feature selection is a critical part of the machine learning workflow. Including features with strong correlation to the target variable (e.g., 'square footage') can improve the model's predictive performance, while features with low correlation (e.g., 'proximity to schools') may not contribute significantly and can be considered for exclusion. Calculating p-values can provide additional insights into the statistical significance of features, helping to make more informed decisions. However, features should not be dropped arbitrarily without proper analysis.
- A. Correct.
Including 'square footage' is valid because a strong positive correlation with the target variable suggests it is likely an important predictor for housing prices.
- B. Incorrect.
This is incorrect because high correlation with the target variable is typically desirable, and there is no indication that this would lead to overfitting in this case.
- C. Correct.
Considering excluding 'proximity to schools' is valid as a low correlation with the target variable suggests it may not contribute significantly to the predictive power of the model.
- D. Correct.
Calculating p-values is an important step to determine the statistical significance of features, which can guide the selection process further.
- E. Incorrect.
This is incorrect because dropping both features without further analysis is premature and does not consider their potential predictive relevance.